TDD Development Workflow

Every feature must be test-driven and E2E validated. See e2e/CLAUDE.md for test standards and failure protocol.

Steps

1. Pick task → bd ready / bd show <id>
2. Create worktree → branch from develop
   bashCopy

   git worktree add ~/worktrees/property-tracker/<name> -b feature/<name> develop
   cp "$CLAUDE_PROJECT_DIR/.env.local" ~/worktrees/property-tracker/<name>/.env.local
   cd ~/worktrees/property-tracker/<name>
   

3. Brainstorm → use context7 for research + design exploration
4. Plan → /feature-dev or write plan manually following project conventions
   • For autonomous execution: follow with /plan-to-epic then /ralph-execute
5. Write tests FIRST (Red) → unit (Vitest) + E2E (Playwright)
6. Spin up env → confirm tests fail (invoke /env-spinup)
7. Implement (Green) → use context7, frontend-design, typescript-lsp, supabase as needed
8. Validate → full env restart, all tests pass
9. Verify → npm run lint && npx tsc --noEmit && npm run build (lint, types, build)
10. PR → gh pr create --base develop → /code-review → gh pr checks --watch → merge
11. Cleanup → remove worktree → bd close <id> → /clear

Writer/Reviewer Pattern (for complex features)

After implementation (step 8), consider opening a second Claude Code session for review:

• Session B reviews with fresh context (no bias toward its own code)
• Use /code-review in Session B
• Address feedback in the original session (Session A)